Representation-Based Exploration for Language Models: From Test-Time to Post-Training

• URL: http://arxiv.org/abs/2510.11686v1
• Date: Mon, 13 Oct 2025 17:49:05 GMT
• Title: Representation-Based Exploration for Language Models: From Test-Time to Post-Training
• Authors: Jens Tuyls, Dylan J. Foster, Akshay Krishnamurthy, Jordan T. Ash,
• Abstract summary: Reinforcement learning (RL) promises to expand the capabilities of language models.<n>It is unclear if current RL techniques promote the discovery of novel behaviors, or simply sharpen those already present in the base model.<n>We investigate the value of deliberate exploration -- explicitly incentivizing the model to discover novel and diverse behaviors.
• Score: 50.144031964319424
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reinforcement learning (RL) promises to expand the capabilities of language models, but it is unclear if current RL techniques promote the discovery of novel behaviors, or simply sharpen those already present in the base model. In this paper, we investigate the value of deliberate exploration -- explicitly incentivizing the model to discover novel and diverse behaviors -- and aim to understand how the knowledge in pre-trained models can guide this search. Our main finding is that exploration with a simple, principled, representation-based bonus derived from the pre-trained language model's hidden states significantly improves diversity and pass@k rates -- both for post-training, and in a novel inference-time scaling setting we introduce. For inference-time, exploration with representation-based diversity improves efficiency, consistently improving pass@k rates across a variety of models and reasoning tasks. For example, for Qwen-2.5-14b-Instruct we obtain over 50% improvement in verifier efficiency on almost all tasks. For post-training, we show that integrating this exploration strategy into an RL pipeline improves reasoning performance over that of the initial model and over standard RL post-training. For example, on AIME 2024, our post-trained Qwen-2.5-7b-Instruct's pass@80 matches the pass@256 of GRPO on the same model, demonstrating a 3x improvement in test-time sample efficiency. Overall, our findings suggest that deliberate exploration -- with the right notion of diversity -- is a practical path toward discovery of new behaviors beyond sharpening.

Related papers

• Outcome-based Exploration for LLM Reasoning [18.33816564983908]
  Reinforcement learning (RL) has emerged as a powerful method for improving the reasoning abilities of large language models.<n>We show that RL can reduce effective diversity even on the training set relative to the base model.<n>We propose outcome-based exploration, which assigns exploration bonuses according to final outcomes.
  arXiv  Detail & Related papers  (2025-09-08T17:52:56Z)
• ExPO: Unlocking Hard Reasoning with Self-Explanation-Guided Reinforcement Learning [12.83211408922535]
  Reinforcement learning-style post-training improves reasoning by optimizing model outputs based on reward or preference signals.<n> GRPO-style approaches implement this by using self-generated samples labeled by an outcome-based verifier.<n>We propose $textbfSelf-Explanation Policy Optimization (ExPO)$-a simple and modular framework that generates such samples by conditioning on the ground-truth answer.
  arXiv  Detail & Related papers  (2025-07-03T17:44:55Z)
• ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models [89.37819814048288]
  We introduce ProRL, a novel training methodology that incorporates KL divergence control, reference policy, and a diverse suite of tasks.<n>Our empirical analysis reveals that RL-trained models consistently outperform base resetting models across a wide range of pass@k evaluations.<n>These findings offer new insights into the conditions under which RL meaningfully expands reasoning boundaries in language models.
  arXiv  Detail & Related papers  (2025-05-30T17:59:01Z)
• RAST: Reasoning Activation in LLMs via Small-model Transfer [33.32587030836428]
  Reinforcement learning (RL) has become a powerful approach for improving the reasoning capabilities of large language models (LLMs)<n>Applying RL at scale remains intimidatingly resource-intensive, requiring multiple model copies and extensive GPU workloads.<n>We propose RAST, a simple yet effective method that transfers reasoning behaviors by injecting RL-induced probability adjustments from a small RL-trained model into larger models.
  arXiv  Detail & Related papers  (2025-05-30T17:57:08Z)
• Echo Chamber: RL Post-training Amplifies Behaviors Learned in Pretraining [74.83412846804977]
  Reinforcement learning (RL)-based fine-tuning has become a crucial step in post-training language models.<n>We present a systematic end-to-end study of RL fine-tuning for mathematical reasoning by training models entirely from scratch.
  arXiv  Detail & Related papers  (2025-04-10T17:15:53Z)
• An Empirical Study on Eliciting and Improving R1-like Reasoning Models [90.52239241349504]
  scaling RL training has become a central technique for implementing such reasoning models.<n>We demonstrate that our RL training approach consistently improves the Qwen2.5-32B base models.<n>We also explore the use of tool manipulation, finding that it significantly boosts the reasoning performance of large reasoning models.
  arXiv  Detail & Related papers  (2025-03-06T15:34:27Z)
• Exploratory Preference Optimization: Harnessing Implicit Q*-Approximation for Sample-Efficient RLHF [82.7679132059169]
  Reinforcement learning from human feedback has emerged as a central tool for language model alignment. We propose a new algorithm for online exploration in RLHF, Exploratory Preference Optimization (XPO) XPO enjoys the strongest known provable guarantees and promising empirical performance.
  arXiv  Detail & Related papers  (2024-05-31T17:39:06Z)
• The Fine Line: Navigating Large Language Model Pretraining with Down-streaming Capability Analysis [27.310894780313618]
  This paper undertakes a comprehensive comparison of model capabilities at various pretraining intermediate checkpoints. We confirm that specific downstream metrics exhibit similar training dynamics across models of different sizes. In addition to our core findings, we've reproduced Amber and OpenLLaMA, releasing their intermediate checkpoints.
  arXiv  Detail & Related papers  (2024-04-01T16:00:01Z)
• SUPERB-SG: Enhanced Speech processing Universal PERformance Benchmark for Semantic and Generative Capabilities [76.97949110580703]
  We introduce SUPERB-SG, a new benchmark to evaluate pre-trained models across various speech tasks. We use a lightweight methodology to test the robustness of representations learned by pre-trained models under shifts in data domain. We also show that the task diversity of SUPERB-SG coupled with limited task supervision is an effective recipe for evaluating the generalizability of model representation.
  arXiv  Detail & Related papers  (2022-03-14T04:26:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.